Can ChatGPT be your coauthor?

In January 2023, the Elsevier journal Nurse Education in Practice ignited a firestorm when it recognized ChatGPT as a coauthor alongside Siobhan O’Connor [Figure].[1] The piece quickly sparked debate among publishers, editors, and researchers about whether a bot can qualify as an author.[2-4]

ChatGPT is an artificial intelligence (AI) language model developed by the company OpenAI. It uses pre-existing books, websites, and other sources to generate human-like text and can assist with things like writing code, composing essays, and answering questions.

Many writers like AI language models because they free up time to focus on higher-level skills like analysis and creativity rather than structure and grammar. Prominent author and Wharton professor Adam Grant has even stated that his classes are now AI mandatory because he does not want to read bad writing anymore.[5] But how should we recognize ChatGPT’s contributions?

According to the International Committee of Medical Journal Editors (ICMJE), the criteria for authorship include four concepts: substantial contributions, drafting the work, final approval, and accountability.[6] The fourth criterion is the most glaringly lacking for ChatGPT, as it cannot be accountable for its work, nor can it assume the moral, legal, and ethical responsibilities required to ensure the accuracy and integrity of the work.[3,6] Furthermore, it cannot own or assign copyright because it does not have legal personality. Some have questioned whether ChatGPT meets any ICMJE authorship criteria, because its contributions are “merely borrowed ideas from its database of information from the Internet,” which is not the same thing as “intellectual content.”[3]

So the verdict, at least for now, is that ChatGPT is not a valid author.[3] Nurse Education in Practice has since published a corrigendum to remove ChatGPT, leaving a sole (human) author on the paper.[8] The most likely approach to AI in publishing appears to be the one taken by JAMA, which is that journals will discourage content created by AI and require a clear description of the contribution.[7] The BCMJ has updated our own author guidelines (https://bcmj.org/submit-article) in accordance with the ICMJE recommendations. However, this consensus[2] among medical journals does not erase the fact that ChatGPT and other forms of AI have become commonplace writing tools and are not going away.

One thing to consider when using ChatGPT in your scientific writing is that it is a content generator, not a reference librarian. OpenAI acknowledges that the technology will generate “plausible-sounding but incorrect or non-sensical answers.”[9] I have found such answers to be frighteningly realistic. For illustration, I asked ChatGPT to create a medical information sheet on “why children should eat ice cream for breakfast.” I told it to include references to published literature and write with a medical lens. Snippets of the result are shown in Boxes 1–3, and as you can see, it both creates references and includes completely fabricated information from this “published literature” in the written text. When I told it that the references were not real, ChatGPT apologized for any confusion and told me that “as an AI language model I do not have the ability to access external sources or verify the accuracy of references.” So, while ChatGPT made a compelling and ostensibly supported argument for ice cream, I am not swapping out my kids’ morning oatmeal just yet.

Ultimately, while AI like ChatGPT can augment human creativity and productivity, the humans behind the AI are responsible for the final interpretation of the work. The debate of AI authorship underlines the importance of understanding the capabilities and limitations of AI, to harness its potential while upholding academic and scientific integrity.
—Caitlin Dunne, MD

hidden

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

References

1.    O’Connor S, ChatGPT. Open artificial intelligence platforms in nursing education: Tools for academic progress or abuse? Nurse Educ Pract 2023;66:103537.

2.    Stokel-Walker C. ChatGPT listed as author on research papers: Many scientists disapprove. Nature 2023;613:620-621. Accessed 5 June 2023. www.nature.com/articles/d41586-023-00107-z.

3.    Teixeira da Silva JA. Is ChatGPT a valid author? Nurse Educ Pract 2023;68:103600.

4.    Sample I. Science journals ban listing of ChatGPT as co-author on papers. The Guardian. 26 January 2023. Accessed 5 June 2023. www.theguardian.com/science/2023/jan/26/science-journals-ban-listing-of-chatgpt-as-co-author-on-papers.

5.    Grant A. TED Audio Collective: ChatGPT did NOT title this podcast. 21 March 2023. Accessed 2 June 2023. www.youtube.com/watch?v=ndQHR3kdEvM.

6.    International Committee of Medical Journal Editors. Defining the role of authors and contributors. Accessed 2 June 2023. www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html.

7.    JAMA Network. Instructions for authors. Last updated 23 May 2023. Accessed 1 June 2023. https://jamanetwork.com/journals/jamaneurology/pages/instructions-for-authors.

8.    O’Connor S. Corrigendum to “Open artificial intelligence platforms in nursing education: Tools for academic progress or abuse?” [Nurse Educ. Pract. 66 (2023) 103537]. Nurse Educ Pract 2023;67:103572.

9.    OpenAI. Introducing ChatGPT. Accessed 31 May 2023. https://openai.com/blog/chatgpt.